A Digital-Twin Driven Modeling Approach of System Degradation for Remaining Useful life Prediction

This paper proposes a digital-twin driven modeling approach for system degradation in remaining useful life (RUL) prediction. We assume the system degradation follows exponential decay, however, the real situation involves complex, nonlinear degradation patterns influenced by various operational factors. A random forest regressor processes both synthetic and real operational data to predict RUL, effectively capturing these degradation dynamics. Results demonstrate accurate modeling of degradation across normal and degraded states. The approach enables predictive maintenance for improving reliability while reducing operational costs through proactive scheduling. This robust, real-time prediction method supports long-duration space mission success.